Negative Bias Temperature Instability (NBTI) is a device degradation mechanism identified in sub-100 nm MOSFETs. When the gate of a P-MOSFET is negatively biased at an elevated temperature, oxide interface traps are generated due to an interaction of holes in the channel inversion layer with gate oxide. The instability is measured as an increase in the magnitude of the threshold voltage of the device. Higher stress temperatures produce more degradation. The mechanism is known to cause reliability performance degradation for the PFET because of this change in threshold voltage. NBTI also results in degradation of transistor drive current for any given drive voltage. NBTI becomes a serious concern for 90 nm technology and beyond, and its effect increases exponentially with decreasing gate oxide thickness.
The term “NBTI lifetime” refers to the period in which the drive current of a transistor at a given drive voltage decreases by 10% from its original value due to NBTI degradation. A transistor that has decreased current output by 10% can have sufficient impact on circuit timing to cause the circuit to fail to perform its intended function, and is thus at the end of its useful life.
A variety of publications have dealt with characteristics, behavior and physics of the NBTI mechanism. Several models have been developed to predict the NBTI lifetime for a given transistor, or circuit containing one or more transistors. NBTI lifetime prediction may in turn be used as part of the system design process, or as part of a qualification procedure for a lot of semiconductor products.
A known NBTI prediction method includes testing transistors at a plurality of different stress voltages (e.g., four different stress voltages) at elevated temperature for a given stress period (e.g., one day at each stress voltage). The transistor drive current at each voltage is monitored throughout the testing. The stress period is long enough so that the drive current of the transistor subjected to the highest stress voltage is reduced by 10%. The transistors subjected to lower stress voltages do not reach 10% degradation during the testing (stress) period, so the amount of time for 10% degradation to occur at each of the lower stress voltages is estimated by extrapolation, based on the drive current degradation measured at each respective stress voltage.
FIG. 1 shows an example in which two different stress voltages are used. The NBTI degradation can be characterized by:ΔId/Id˜tn,  (1)
where Id is the drive current, t is the length of time the stress voltage is applied, and n is a parameter referred to herein as the degradation slope.
At a stress voltage level of −2.0V, the transistor reaches 10% degradation, and the stress testing period is then ended. During the same stress period, the transistor subjected to −1.4 V does not reach 10% degradation, so the length of time to reach 10% degradation at 1.4 V is estimated by extrapolation. As shown in FIG. 1, the extrapolated estimate of stress time for the −1.4 V stress voltage can differ substantially depending on whether the extrapolation is based on the full range of measured degradation values, or only on the last five points measured.
There has been controversy regarding these models and the selection of data points to be used in the models. For example, consider a first sample including the full range of drive current data points collected throughout the stress period at one of the lower voltages, such that drive current degradation during the stress period is less than 10%. Also consider a second sample including only a subset of the same data in the first sample (e.g., the five most recently collected drive current data points at the end of the stress period). Using a conventional NBTI model, the predicted NBTI based on the first sample may differ from the predicted NBTI based on the second sample (because of the extrapolation used to determine the stress time for the lower voltage levels). For example, the predicted NBTI based on the full range of data from the first sample may be substantially shorter than the predicted NBTI based on the subset of data in the second sample.
FIG. 2 shows an example of NBTI predictions based on stress testing at four different stress voltages: −1.1, −1.55, −1.7 and −1.85 Volts. The line indicated by ellipses is based on the stress time extrapolated based on the full range of test data. The line indicated by squares is based on the stress time extrapolated based on the last five test data. The results based on the full data range at each stress voltage predict a shorter NBTI lifetime than the results based on the last five data points at each stress voltage.
Neither the extrapolation based on the full range of sample selection or the extrapolation based on the last-five-points has achieved universal acceptance.
A model that underestimates the NBTI lifetime is troublesome, because it causes over-design of circuits and increases the cost of products that are required to last for a given specified product life.